Collected sources and patterns will appear here. Add from search or the patterns library.
DeviceTensor -> FrameworkTensor
Share underlying GPU device memory pointers directly with DL frameworks to construct native tensors without copying data.
Problem it solves
Exporting preprocessed tensors to training frameworks via host memory round-trips degrades performance.
Consumes
Emits
The real projects this mechanism was found in. Attribution is the point — this is how the best teams actually do it.